After creating YOLO v2 with resnet50, while training the network, getting error "Invalid training data. The output size ([7 7 16]) of the last layer does not match the response size ([1 1 3])."

Question

Susama Bagchi 2019-12-9

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/495651-after-creating-yolo-v2-with-resnet50-while-training-the-network-getting-error-invalid-training-da

编辑： Susama Bagchi 2019-12-22

I followed the YOLO v2 object detection network as given in the below link:

https://in.mathworks.com/help/vision/ref/yolov2layers.html

Then I used the imagedatastore and augmented image datastore to modify the image size. After that, when I tried to train the network, got the error "Invalid training data. The output size ([7 7 16]) of the last layer does not match the response size ([1 1 3])."

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Sourav Bairagya 2019-12-18

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/495651-after-creating-yolo-v2-with-resnet50-while-training-the-network-getting-error-invalid-training-da#answer_406868

From the error it seems that output of your network should have size [1 1 3], but it is having the size [7 7 16]. May be there is some error occured in the training input images during modification.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Susama Bagchi 2019-12-22

编辑：Susama Bagchi 2019-12-22

If I only use the resnet50, with modified image size also it is working. But when I am adding YOLO V2 at the end of resnet50 as per the link, I am getting error. Even if I don't change the input layer size, then also output size mismatch error is there.

I hope, I can train the network after adding YOLO V2. If any other option, please let me know.

请先登录，再进行评论。